Approved Research

Tools for evaluating bias in ADRD research

Lay summary

Critical social exposures and potentially modifiable risk factors for Alzheimer's disease and dementia occur across the lifespan, prior to the diagnosis of these conditions. For example, Alzheimer's disease and dementia is associated with social identities such as gender, education and income (and other socioeconomic factors), race, and ethnicity. In addition, a number of potentially modifiable behavioral or physiologic characteristics are associated with increased risk of Alzheimer's disease and dementia.

However, various biases make it difficult to accurately assess the effects genetic, social, and modifiable risk factors. The ultimate goal of our research is to pinpoint how and when we can intervene to prevent or delay the onset of Alzheimer's disease and dementia. However, a socially and biologically informed approach that incorporates the effects of potential biases will be required to accomplish this. We propose rigorously evaluating the effects of genetic, lifecourse social, and modifiable risk factors using a diverse set of epidemiologic methods and approaches to evaluate, quantify, and correct for potential biases.

This three-year study will improve the validity of lifecourse research cohorts and provide more valid and public health relevant estimates of the effects of genetic, social and potentially modifiable determinants of Alzheimer's disease and dementia. The tools will be useful to other Alzheimer's disease and dementia researchers. UKBiobank provides an incomparable source of data because of its size, the detailed characterization of the cohort, and the heterogeneity of participants compared to the other cohorts we plan to leverage.

Scope extension:

To what extent does unmeasured confounding, differential survival, misclassification of exposures, enrollment, and attrition and reverse causation bias observed associations between genetic, social, and modifiable risk factors for Alzheimer's disease and related dementias (ADRD)?

We propose the following four aims:

1) To evaluate confounding, differential survival, misclassification, enrollment, and attrition and other factors that may bias estimated effects of genetic, social, and modifiable risk factors for ADRD. For example, prior research indicates misclassification of risk factors such as educational attainment, sex/gender, and physical activity is common and may bias estimated associations with ADRD.

2) To evaluate reverse causation-in which incipient dementia induces changes in risk factors-by using instrumental variable (IV) analyses, including Mendelian Randomization/genetic IVs and policy-based IVs.

3) Using results from 1 and 2, to rigorously estimate the causal effects of genetic, social, and modifiable risk factors on ADRD risk using both UKBiobank alone and a synthetic cohort of pooled cohorts corrected for misclassification, selection, survival, and reverse causation biases.

4) To quantify reduction in lifetime ADRD cases and ADRD gender and racial disparities that could be achieved with a variety of hypothetical interventions on social or vascular or other potentially modifiable risk factors at different ages.

New Scope:

We are publishing a paper on gender misclassification and enrollment differences. This project falls under Aim 1, which states we evaluate will evaluate confounding, differential survival, misclassification, enrollment, and attrition for various ADRD risk factors. Gender is a strong determinant of ADRD, disproportionately affecting women due to differences in life expectancy, hormonal and other biological factors, medication use differences (e.g. HRT), and social factors. Our study quantifies the disagreement between chromosomal and self-reported sex and identifies potential reasons for discordance using data from the UK Biobank. Most disagreement observed appears to indicate medically and demographically meaningful information that is unlikely to be due to typographical or self-reporting errors. This work lays the groundwork for correcting associations between gender and dementia due to misclassification and better understanding medical and social determinants of dementia risk, especially endogenous and exogenous hormonal determinants, in future work.